Pytorch implementation of "Parallax Attention for Unsupervised Stereo Correspondence Learning", TPAMI 2020

[arXiv]

• Python 3.6
• PyTorch >= 1.1.0
• prefetch_generator

Download SceneFlow and KITTI 2015 datasets.

Run ./train.sh to train on the SceneFlow dataset. Please update datapath in the bash file as your training data path.

Run ./finetune.sh to finetune on the KITTI 2015 dataset. Please update datapath in the bash file as your training data path.

Download pre-trained models to ./log.

• Google Drive
• Baidu Drive[code:fe12]

Run ./test.sh to evaluate on the test set of the SceneFlow dataset. Please update datapath in the bash file as your test data path.

Run ./submission.sh to save png predictions on the test set of the KITTI 2015 dataset to the folder ./results. Please update datapath in the bash file as your test data path.

@Article{Wang2020Parallax,
  author    = {Longguang Wang and Yulan Guo and Yingqian Wang and Zhengfa Liang and Zaiping Lin and Jungang Yang and Wei An},
  title     = {Parallax Attention for Unsupervised Stereo Correspondence Learning},
  journal   = {{IEEE} Trans. Pattern Anal. Mach. Intell.},
  year      = {2020},
}

This code is built on GwcNet. We thank the authors for sharing their codes.